Prediction and Optimization of Heavy Clay Products Quality - Advanced Materials for Agriculture, Food and Environmental Safety - page 94

Trapezoidal membership function used in this calculation could be
written as:

xa −

axm ma ≤< ,

−

Axamnb (,, ,,) = m x n ≤ < , 1 (4.5)

xn −

nxb ≤< , 1 − −

bn

Where x is whether CS , WA , FS , WLF , and VMC , and the values of a , b , m ,
and n are function parameters. Interval a - b represents the range in which
measured values occurred, while range m - n is the expected optimal value
range for output variables, chosen for certain product groups.

4.3 Results and Discussions

Experimental data obtained (chemical composition and physical-mechan-
ical characteristics) were presented using basic descriptive statistics of
obtained data, in table 4.1. Both chemical content and i nal products char-
acteristics varied signii cantly, implying that i tting of the experimental
data could be performed using SOP and ANN modeling.
h e 139 heavy clay samples showed the expected typical chemical com-
positions, rich in silica and aluminum, accompanied by relatively signii -
cant amounts of iron oxides, with satisfying content of potash, and minor
contents of titan. Harmful constituents that can cause el orescence were
not detected in most of the samples [21-23]. Large dif erences among tech-
nological characteristics indicated that the tested samples contained vari-
ous contents of clay and other minerals.

4.3.1 Correlation Analysis

h e correlation coei cients between independent and dependent parame-
ters in this research are presented in table 4.2. Most of the coei cients were
statistically signii cant (p<0.01), due to the large number of measurements.
h e analysis is done to test the parameters prior to further investigation
of the models. Since correlations presume linear relationships [24], the
obtained coei cients were not very high, and the non-linearity of the sys-
tems was indicated. h e highest positive correlations were observed in the